Project Summary Animals are constantly exposed to a barrage of multisensory input from their stimulus-rich environments. They handle this informational complexity by having their behavior guided by the most physically salient (or more generally, the most important) stimulus source in the environment. The identification of the most physically salient stimulus occurs through neural mechanisms of stimulus competition, which must necessarily operate across sensory modalities and across spatial locations. Although the mechanisms of multisensory integration have been studied extensively, the circuit and computational principles underlying competition within and across sensory modalities are largely unknown. Recent evidence from behaving monkeys has revealed the midbrain superior colliculus (SC) as being critical for normal competitive stimulus selection. In parallel, our recent work in the barn owl optic tectum (OT, the avian homolog of the SC) has revealed special neural response properties, namely categorical signaling of the strongest stimulus, that can account for the SC's critical role in selection behavior. Inhibition from a GABAergic midbrain nucleus, the isthmi pars magnocellularis (Imc), is necessary to mediate these response properties. Nonetheless, the computational and mechanistic logic of Imc function in service of competitive stimulus selection remain unknown. Here, we propose to systematically unravel fundamental computations orchestrated by the Imc-OT network for multisensory competition, and to map their implementation explicitly onto circuit elements. Specifically, we first aim to elucidate how the reliable signaling of the strongest stimulus in the presence of noise, i.e, ?robust? signaling, is implemented. Our hypothesis is that special donut-like patterns of spatial inhibition from the Imc to the OT play a central role. Second, we aim to understand if the Imc is an active computational locus for stimulus competition in the OT. Our hypothesis is that competitive interactions within the Imc control the accuracy and strength of categorization by the OT. Third, we ask how the OT resolves competition in cluttered sensory scenes that contain several stimuli. Our hypothesis is that a dynamic inhibitory balance among the multiple competing locations protects OTid responses from being driven to zero and permits network wide decoding of the strongest stimulus. We will test the hypotheses using in vivo electrophysiology and drug iontophoresis in awake, head-fixed barn owls together with computational modeling. In all cases, we will explicitly test whether the hypothesized mechanisms of competition generalize across sensory modalities. Preliminary data from the three aims support our hypotheses. They indicate that results from the proposed experiments have the power to reveal strategic principles of circuit organization for executing the sophisticated computations that subserve multisensory competition and stimulus selection.